A Probability-Based Framework for Dynamic Resource Scheduling in Grid Environment

Abstract

Recent enthusiasm in grid computing has resulted in a tremendous amount of research in resource scheduling techniques for tasks in a workflow. Most of the work on resource scheduling is aimed at minimizing the total response time for the entire workflow and treats the estimated response time of a task running on a local resource as a constant. In this paper, we propose a probabilistic framework for resource scheduling in grid environment that views the task response time as a probability distribution to take into consideration the uncertain factors. The goal is to dynamically assign resources to tasks so as to maximize the probability of completing the entire workflow within a desired total response time. We propose three algorithms for the dynamic resource scheduling in grid environment. Experimental results using synthetic data derived from a real protein annotation workflow application demonstrate that considering the uncertain factors of task response time in task scheduling does yield better performance, especially in a heterogeneous environment. We also compare the relative performance of the three proposed algorithms.